English

PAC-Chernoff Bounds: Understanding Generalization in the Interpolation Regime

Machine Learning 2025-02-11 v4 Statistics Theory Machine Learning Statistics Theory

Abstract

This paper introduces a distribution-dependent PAC-Chernoff bound that exhibits perfect tightness for interpolators, even within over-parameterized model classes. This bound, which relies on basic principles of Large Deviation Theory, defines a natural measure of the smoothness of a model, characterized by simple real-valued functions. Building upon this bound and the new concept of smoothness, we present an unified theoretical framework revealing why certain interpolators show an exceptional generalization, while others falter. We theoretically show how a wide spectrum of modern learning methodologies, encompassing techniques such as 2\ell_2-norm, distance-from-initialization and input-gradient regularization, in combination with data augmentation, invariant architectures, and over-parameterization, collectively guide the optimizer toward smoother interpolators, which, according to our theoretical framework, are the ones exhibiting superior generalization performance. This study shows that distribution-dependent bounds serve as a powerful tool to understand the complex dynamics behind the generalization capabilities of over-parameterized interpolators.

Keywords

Cite

@article{arxiv.2306.10947,
  title  = {PAC-Chernoff Bounds: Understanding Generalization in the Interpolation Regime},
  author = {Andrés R. Masegosa and Luis A. Ortega},
  journal= {arXiv preprint arXiv:2306.10947},
  year   = {2025}
}

Comments

60 pages, 12 figures, published at JAIR 2025

R2 v1 2026-06-28T11:08:47.410Z